LM317AHVT and Heatsink

[mike_mike]

Dear Forum,

I built the power supply from the attachment (with a few modifications: protection diodes and a capacitor in parallel with the potentiometer) and I mounted on one heatsink the power transistor and on another heatsink the rectifier and the LM317AHVT (I will note LM317 for LM317AHVT from now)
The heatsink with LM317 is having the following dimensions: 165x80x35mm.
I mounted the LM317 using a very thin piece of mica and I used thermal grease and a M3 screw and a mounting clip.
The problem is that I noticed that LM317 is heating up to about 70 C while on the power supply output is connected a overload.
The current through the LM317 is 0.7A and the voltage drop on the LM317 is 33V, while on the output is an overload.
It is a problem if LM317 is heating up to about 70C after a few minutes of overload ?
It is normal to heat up to 70 C or this temperature is too high ?

[james_s]

That sounds pretty reasonable, with 33V on the input you are dissipating quite a lot of power. Is the heatsink getting hot too? If the IC is running much hotter than the heatsink I would look at the mounting to make sure it is properly mated. If the heatsink is getting hotter than you like then use a larger heat. There is no getting around a linear regulator dropping a large voltage at significant current getting hot, you can't fight the math.

[Zero999]

70^oC isn't too hot.

Another thing to note is the safe operating area protection is kicking in and reducing the current limit. When the voltage differential between the input and output exceeds 15V, the output current limit is reduced, to protect the pass transistor from secondary breakdown. With a voltage differential of 60V, the current is typically limited to 300mA.
https://docs-emea.rs-online.com/webdocs/12cf/0900766b812cfcaa.pdf

The solution is to either route more current through the pass transistor, or add another transistor in series with the LM317HV, to reduce the voltage across it.

[mike_mike]

Thanks for the answers.
James, the heatsink is getting hot after about 30 mins of overload. The LM317 is properly mated (I think). I used a very thin piece of mica insulator  (about 0.05mm thickness).

The temperature was measured using a thermocouple and was the case temperature, not the junction temperature.
In this case, your answers are still the same ?

[james_s]

What is the heatsink temperature?

70C case temperature doesn't sound unreasonable, it would worry me if it was above about 85-90C, but 70 is within the range that is acceptable.

[not1xor1]

According to the datasheet (the first I found is from Fairchild) the worst case minimum load for stability is 12mA so R1 should be around 100 ohm and so a 2.2k potentiometer should be used for P1.

In many cases LM317s would work stably with a much lower load than 12mA, but if I were you, I would test the IC on a breadboard rather than assuming to be always kissed by the blind Luck goddess  .

[mike_mike]

James, I did not measure the temperature of the heatsink, but I know that if I touched the heatsink it burned me after 2-3 seconds.
not1xor1, I did not mentioned, but I used a 1.2k resistor on the board as a load. I can change this resistor to a more suitable value of resistance.

[james_s]

Well so long as the case is not hugely hotter than the heatsink then there is not much you could do besides use a larger heatsink or forced air. Linear regulators turn the excess voltage into heat, they run hot, no way to change that.

[schmitt trigger]

I would be far more concerned with the PNP series pass transistor overheating, as on its own it does not have thermal shutdown.

Many designs thermally couple the transistor with the LM317, such that the heat generated by the former will indirectly cause the LM317's thermal shutdown to kick in.
Works fine if the thermal coupling is quite tight.

[Zero999]

I would be far more concerned with the PNP series pass transistor overheating, as on its own it does not have thermal shutdown.

Many designs thermally couple the transistor with the LM317, such that the heat generated by the former will indirectly cause the LM317's thermal shutdown to kick in.
Works fine if the thermal coupling is quite tight.

True, at least this design does have current limiting, thanks to D1, D2 and R6.

[mike_mike]

I tried to modify the schematic in order to have a minimum load current of about 10mA.
I changed the 430R resistor into a 120R resistor. I used the same potentiometer (10K). I connected a 3.3K resistor in parallel with the potentiometer to have a variable output voltage from 1.25V to about 26V.
I tried the schematic and it works.
Are those modifications correct ?

[mikerj]

I changed the 430R resistor into a 120R resistor. I used the same potentiometer (10K). I connected a 3.3K resistor in parallel with the potentiometer to have a variable output voltage from 1.25V to about 26V.
I tried the schematic and it works.

Having a relatively low value fixed resistor in parallel with the variable resistor gives very non-linear behaviour of the voltage control e.g. for a standard 270 degree potentiometer the first 90 degrees would  change the voltage by about 17v, but the last 90 degrees would only change the voltage by about 3v. 

This may be acceptable (or even desired) behaviour, but if you want more linear behaviour use a lower value potentiometer and a higher value fixed resistor to set the maximum voltage (e.g. 2k7 pot, 27k resistor)

[mike_mike]

Can this modification (from my last message) lead to unwanted power supply behavior, such as oscillations ?

[mikerj]

Can this modification (from my last message) lead to unwanted power supply behavior, such as oscillations ?

No, the modifications should not affect stability.  Reducing the upper feedback resistor to 120 ohms should improve voltage regulation with small loads.

[JS]

I don't know about the 1.2Ω resistor, seems a tad high, if you lower that value more current will go by the transistor and less for the regulator to take. I don't know your current requirements, at 3A (taking 0.65V for VBE) the transistor will be taking 2.6A and the reg 0.4A. If you lower it to 0.47Ω the transistor will take 2.8A and the reg 0.2A...

JS

[mike_mike]

Thanks for the replies.

I have not found 120R resistor at the local electronic components shop. So I used 100R in series with 22R. I calculated that the current for 120R resistor is 1.25V/120R=0.0104A, while using the 122R the current is 1.25V/122R=0.0102A. The difference of current is very small.
It should be ok if I use 100R in series with 22R ?

[Zero999]

122R vs 120R should make no practical difference, since it's less than the tolerance of the LM317.

[newbrain]

I don't know about the 1.2Ω resistor, seems a tad high, if you lower that value more current will go by the transistor and less for the regulator to take.

Yes, I also have some doubt about the emitter resistor value; as it is, I expect the current through the transistor not to exceed about 1 A.
Reasoning:
We have two diodes drop (D1+D2) voltage across the resistor R6 and the BE junction of Q1.
Let's say it's about 1V (as per the DS, with 3A through the diode), so 2V total (excess rounding).
Across R6 we have 2 - 0.7 = 1.3V.
Ohm's law gives us a current of 1.3V/1.2 = ~1.1A

All the rest will be carried by the regulator.

...at 3A (taking 0.65V for VBE) the transistor will be taking 2.6A and the reg 0.4A. If you lower it to 0.47Ω the transistor will take 2.8A and the reg 0.2A...

Am I missing something? See above my calculations.

[JS]

...at 3A (taking 0.65V for VBE) the transistor will be taking 2.6A and the reg 0.4A. If you lower it to 0.47Ω the transistor will take 2.8A and the reg 0.2A...

Am I missing something? See above my calculations.

You are right, I'm ignoring the diodes, as if they were just to limit the maximum current as you said but still working on he linear mode, my mistake. Reducing the value of the emmiter resistor will send more current to the transistor even if the diodes are conducting. 1.4V at the diodes, -0.65 VBE, 0.75V/0.47Ω gives 1.6A to the tranny, but limiting the current to the reg at 140mA for resistive share between the transistorand reg. I used 0.2A at the reg resistor, which means 2V before conducting by the diodes, so using 3 diodes instead of 2 makes that work, to keep the 2 diodes the emmiter resistor needs to be even lower and the approximations of VBE and Vf are more on the limit, where you probably want to measure them at the limit conditions to know how much current will be shared.

JS

[mike_mike]

It should be a good idea to use a 1K potentiometer in series with the 10K potentiometer for fine adjustment of the output voltage ?
Will the power supply work correctly if I use the 1K potentiometer in series with the 10K potentiometer ?

[JS]

No problem with dual pots, you might want to use a 500Ω instead, as it might give ypu better control over the range, but 1k should be fine as well.

JS

[mike_mike]

Dear Forum,

I made a layout for a LM317 power supply, using the attached schematic.
I don't know if I correctly made the layout. Please have a look at the layout and tell me if it is good.
I want to know only if the layout is correct. The schematic that I draw in KiCAD is correct because I checked it. I also added a LED and a current limiting resistor (R4 and J3).
The refdes of the components does not correspond from the schematic to the layout.
I was a little bit confused by the copper pours, because I don't know if I correctly made the copper pours.

[mike_mike]

I have a question regarding the attached schematic.
In the LM317AHVT's datasheet is noted that the minimum load current can vary between 3.5mA and 12mA. When calculating the value of R1, what value 3.5mA or 12mA should I use ?
The current value of R1 is 122ohms (not 120 ohms), and the minimum load current in this situation is 10.2mA. It is sufficient this value of current for stabilization or it should be higher ?

[Zero999]

I have a question regarding the attached schematic.
In the LM317AHVT's datasheet is noted that the minimum load current can vary between 3.5mA and 12mA. When calculating the value of R1, what value 3.5mA or 12mA should I use ?
The current value of R1 is 122ohms (not 120 ohms), and the minimum load current in this situation is 10.2mA. It is sufficient this value of current for stabilization or it should be higher ?

Well spotted. There's a lot of variance between parts with the same number. With 10.2mA, there's the very tiny outside chance the voltage may rise slightly beyond the specification, when it's unloaded, especially at extreme operating conditions i.e. high voltages high/low temperatures.

How many of these are you going to build? 99.99% of parts will work fine with 10.2mA, especially if their operating conditions aren't being pushed to extremes.

[mike_mike]

I already built 2 power supplies and I am planning to build the 3 rd one.
I also planning a mini series (10 pieces) with PCB's from China.